Abstract
Context
For patients with metastatic cancer and limited life-expectancy, potential benefits of chemotherapy must be balanced against harms to quality-of-life near death and increased out-of-pocket costs of care.
Objectives
To evaluate the association between chemotherapy use by patients with Stage IV pancreatic cancer and health care use and Medicare and out-of-pocket costs in the last 30 days of life.
Methods
We conducted a retrospective, cohort study of 3,825 patients 66 years or older when diagnosed with Stage IV pancreatic cancer in 2006–2011, using the linked Surveillance, Epidemiology, and End Results (SEER)-Medicare data. Using propensity-score matched sample, we examined associations between initiation of chemotherapy shortly after the metastatic diagnosis (and secondarily, continued chemotherapy use in the last 30 days of life) and healthcare use and costs (both Medicare payment and patient out-of-pocket costs) in the last 30 days of life.
Results
Chemotherapy use was associated with increased rates of hospital admissions (45.0% vs. 29.2%, p<0.001), emergency department visits (41.3% vs. 27.2%, p<0.001), and death in a hospital (14.2% vs. 9.1%, p<0.001), fewer days in hospice care (11.5 days vs. 15.7 days, p<0.001), and more than 50% increase in patient out-of-pocket costs for care ($1,311.5 vs. $841.0, p<0.001) in the last 30 days of life. Among patients who initiated chemotherapy, more stark differences in these outcomes were found by whether patients received chemotherapy in the last 30 days of life.
Conclusion
Chemotherapy use among older patients diagnosed with metastatic pancreatic cancer was associated with substantially increased use of healthcare and higher patient out-of-pocket costs near death.
Keywords: chemotherapy, healthcare use, end-of-life, out-of-pocket costs, financial burden
Introduction
Patients diagnosed with metastatic cancer often must choose between potential benefits of chemotherapy (e.g., prolonged survival; symptom relief) and its potential harms to quality-of-life near death (QOD). On one hand, chemotherapy is associated with modest life-extending benefits for selected types of cancer. For example, a median survival benefit of about 30 days was found for older patients diagnosed with Stage IV non-small cell lung cancer in routine practice.(1,2) On the other, chemotherapy use in the last 6 months of life among patients with metastatic cancer that has progressed beyond initial treatment was associated with increased use of intensive, life-prolonging care in the last week of life, late referral to hospice, and a greater likelihood of dying in an intensive care unit and lower likelihood of dying at home.(3) Additionally, recent evidence suggests that such chemotherapy use was associated with worse QOD among patients with good performance status and no improvement among patients with poor performance.(4) Receipt of aggressive anticancer care in intensive care settings in patients near death was also associated with worse family evaluation of care for their loved ones.(5)
Further complicating the tradeoffs of chemotherapy use in advanced cancers are the enormous societal and personal costs associated with cancer care at the end of life. Roughly 25% of Medicare costs go to care in the final year of life,(6) with about 40% of this expense (or 10% of the overall Medicare budget) spent in the last month of life.(7) Notably, patients with cancer account for 22% of the top 5% beneficiaries of Medicare, representing a significant cost burden.(8) Patients with advanced cancer and their families are at a high risk of medical debt and bankruptcy as a result of substantial out-of-pocket (OOP) costs of treatment,(9,10) which, in turn, is a major stressor at the end-of-life and associated with worse quality-of-life.(11,12)
Pancreatic cancer is the 4th leading cause of cancer deaths in the U.S. and projected to become the second leading cause by 2020.(13) Because metastatic pancreatic cancer patients have a limited life-expectancy, there is an ethical interest in ensuring that patients are aware of chemotherapy’s potential harms to QOD so that they can make informed choices among end-of-life care options.(14,15) Two recent clinical trials enrolling metastatic pancreatic cancer patients with good performance status found a median survival from diagnosis to be 8–11 months for patients receiving combination chemotherapy and 6–7 months for patients with single-agent treatment.(14,15) Prognosis for patients with poorer performance is likely worse. (To our knowledge, no clinical trial compared chemotherapy with no chemotherapy in this population.) The very limited life-expectancy and the generally low response of pancreatic cancer to chemotherapy, especially second-line treatment,(16,17) heightens the value of treatment approaches aimed at palliation and improvement of QOD and avoidance of care and death in an intensive care setting. This study aims to assess the association between chemotherapy use and health care utilization and costs close to death among patients diagnosed with metastatic pancreatic cancer. We examined such associations for both the initiation of chemotherapy shortly after a diagnosis of metastatic pancreatic cancer and additional chemotherapy use in the last 30 days of life. We hypothesize that initiation of chemotherapy (and continued chemotherapy use in the last 30 days of life) is associated with a greater likelihood of receiving health care in hospitals or emergency departments (EDs), lower use of hospice care, and greater financial burdens in term of both Medicare payment and patient OOP costs, in the last 30 days of life.
Methods
Sample and Data
The sample included patients aged 66 or older at the time of their metastatic (Stage IV) pancreatic cancer diagnosis and who died within 12 months of their diagnosis. We used data from the SEER-Medicare linked database with dates of diagnosis from 2006–2011 and excluded patients for whom the metastatic pancreatic cancer diagnosis was not the first primary cancer diagnosis and patients diagnosed with any other cancer. To derive baseline co-morbidity indicators and capture healthcare use and costs close to death, we required continuous enrollment in fee-for-service Medicare Parts A&B for at least 12 months prior to and 12 months post diagnosis (or until death if earlier). We further restricted our sample to patients who survived until at least 60 days post-diagnosis, excluding those for whom chemotherapy might not have been offered because of poor prognosis or performance status.(1) We also required chemotherapy, if taken, to be initiated within 60 days of diagnosis, to focus on treatment decisions shorty after diagnosis.(1)
Measures
Chemotherapy use was identified by applying ICD-9 diagnosis and procedure codes and HCPCS codes for chemotherapy administration or drugs(18) to Medicare claims reflecting care in physician offices, outpatient clinics, and inpatient settings. All unique days on which a given patient received chemotherapy from 30 days prior to diagnosis (to allow for discrepancies between actual date of service and date as billed) until death were captured. We then determined whether chemotherapy was initiated within 60 days of diagnosis, and, among those who initiated chemotherapy, whether the patients received chemotherapy in the last 30 days of life.
End-of-life health care utilization and costs were measured over the last 30 days of life and included: dichotomous indicators of having had at least one hospital admission, at least one ED visit, whether the patient received any hospice care, and whether the patient died at a hospital (vs. another place such as home); number of days of hospice care received; Medicare payment for all Medicare-covered services received in the last 30 days of life, patient OOP costs, and the sum of Medicare payment and patient OOP costs, all adjusted to 2012 dollars using the Consumer Price Index.(19) (see eTable 1 for details).
Patient age, gender, race/ethnicity, marital status, foreign-born status, urban/rural status of residence according to US Census categorization,(20) were documented. Clinical characteristics included number of co-morbid conditions derived using algorithms for Charlson comorbidity indices provided by the National Cancer Institute,(21) and whether the patient had surgery or radiation as part of initial course of treatment. Neighborhood socioeconomic indicators used in the analysis included Census Tract level median household income, % of residents living below the federal poverty line, and % of residents 25 years or older with < 12 years of education. We categorized all Census Tracts nationwide into quartiles and used quartile indicators in our analyses.
Statistical Analysis
Bivariate analyses examined patient baseline characteristics by receipt of chemotherapy, and, among patients who received chemotherapy, by whether or not patient had chemotherapy in the last 30 days of life.
Propensity scores(22) were constructed to achieve balances between the treatment (chemotherapy) and comparison (no chemotherapy) patients within each block with regard to patient demographics (gender, age, race/ethnicity, marital status), clinical characteristics (number of co-morbid conditions, surgery or radiation as part of first course of treatment, and neighborhood socio-economic indicators. Year of diagnosis was not included because patients with and without chemotherapy had almost identical distribution in years of diagnosis. We used 1:1 nearest-neighbor matching within a defined caliper (.25 of standard deviation of the log-odds of estimated propensity scores)(22,23) with replacement, which allowed patients not receiving chemotherapy to serve as matches in more than one pair.
Using the matched sample, we compared healthcare use and costs by chemotherapy use and tested for differences using Chi-square tests for dichotomous outcomes and median tests for count and continuous outcomes. We also examined median survival by chemotherapy use in our sample but caution against interpreting the results as “survival benefits” of chemotherapy because, first, the confounding by (unobserved) performance status and prognosis is likely to overstate the survival benefits of chemotherapy; second, our retrospectively defined sample – patients who died within 12 months of diagnosis – may lead to underestimation. Together, these biases render a true prospective survival analysis not possible.
Using the propensity score-matched sample,(24) we estimated a logistic model for each dichotomous outcome (any hospital admission, ED visit, hospice use, and death in a hospital). We estimated a Zero-inflated Poisson (ZIP) regression for the number of days receiving hospice care (including zero days) in the last 30 days of life; the ZIP model explicitly models the process of having any hospice care separately from the process of days in hospice conditional on having at least one day of care.(25) For Medicare payment, patient OOP costs, and total payment (Medicare + OOP), we estimated a generalized linear model with a log link function and gamma distribution.(26)
Propensity scores can be used to match or weight samples in a variety of ways in order to adjust for observed confounding, but no one method is optimal across all analyses.(27) To ensure that our results were not sensitive to choice of propensity score adjustment strategy, we repeated our analyses with regressions using inverse probability weights,(27) in which treated individuals are assigned a weight of the inverse of the propensity score and comparison individuals are assigned a weight equal to the inverse of one minus the propensity score. In all analyses, we derived predicted outcomes and marginal effects associated with chemotherapy and present estimated average treatment effects (ATEs) for the entire sample. We did not further adjust the standard errors of the marginal effects to account for uncertainties in propensity score estimation because ignoring such uncertainties is associated with overestimates of standard errors.(28)
Among patients with chemotherapy use, baseline patient characteristics were largely balanced between those who had chemotherapy in the last 30 days of life and those who did not (with all standardized differences<10%). We followed the same analytical strategies for this comparison but without propensity score matching or weighting. All analysis was conducted using SAS 9.4 and STATA 14.0.
This study was approved by the Institutional Review Board of the Weill Cornell Medical College, which determined that informed consent was not required.
Results
Chemotherapy vs. no chemotherapy
Our study sample contained 3,825 patients diagnosed with Stage IV pancreatic cancer during 2006–2011 who survived until at least 60 days but died within 12 months of diagnosis; 1,840 patients initiated chemotherapy within 60 days of diagnosis (48.1%); 1,985 patients did not have any chemotherapy. A total of 615 patients were excluded from the sample because their first chemotherapy took place after 60 days from diagnosis. Patients with chemotherapy (vs. without) were more likely to be male, younger, non-Hispanic whites (vs. racial/ethnic minorities) and married (p<0.001 for all comparisons), and were less likely to have 3 or more Charlson comorbidities and much less likely to have had surgery as part of initial course of treatment than patients who did not receive chemotherapy. Patients receiving chemotherapy were more likely to live in neighborhoods with higher median household income, lower level of poverty, and lower level of high-school incompletion than patients who did not receive chemotherapy (Table 1).
Table 1.
Baseline patient characteristics for entire sample and by chemotherapy use
| Baseline characteristics | Entire Sample N=3,825 |
Chemo = 0 N=1,985 |
Chemo = 1 N=1,840 |
p - value |
|---|---|---|---|---|
| Female | 55.1 | 59.1 | 50.7 | <0.001 |
| Age at diagnosis | ||||
| 66–69 | 18.4 | 12.4 | 24.9 | |
| 70–74 | 19.6 | 15.0 | 24.5 | <0.001 |
| 75–79 | 28.0 | 26.5 | 29.7 | |
| 80+ | 34.0 | 46.1 | 20.9 | |
| Race/ethnicity | ||||
| White, non-Hispanic | 76.4 | 72.7 | 80.5 | |
| Black, non-Hispanic | 10.5 | 12.5 | 8.4 | <0.001 |
| Hispanic | 6.3 | 6.8 | 5.7 | |
| Other | 6.8 | 8.0 | 5.4 | |
| Marital status | ||||
| Married | 51.4 | 42.6 | 60.9 | |
| Never married | 7.8 | 8.6 | 7.1 | |
| Separated or divorced | 8.5 | 8.9 | 8.1 | <0.001 |
| Windowed | 29.3 | 36.9 | 21.0 | |
| Marital status unknown | 2.9 | 3.0 | 2.9 | |
| Urban/rural | ||||
| Big metro | 56.5 | 57.7 | 55.3 | |
| Metro | 28.2 | 27.5 | 29.0 | 0.456 |
| Urban | 5.7 | 5.3 | 6.0 | |
| Less rural or rural | 9.6 | 9.5 | 9.7 | |
| Foreign born | 30.3 | 30.1 | 30.4 | 0.864 |
| # of Charlson comorbidities | ||||
| 0 | 45.8 | 43.5 | 48.3 | |
| 1 | 28.7 | 27.5 | 30.1 | <0.001 |
| 2 | 13.5 | 14.7 | 12.1 | |
| 3 or more | 12.0 | 14.4 | 9.5 | |
| Year of diagnosis | ||||
| 2006 | 16.3 | 16.6 | 16.1 | |
| 2007 | 17.3 | 18.0 | 16.5 | |
| 2008 | 16.7 | 16.1 | 17.3 | 0.224 |
| 2009 | 16.4 | 15.7 | 17.1 | |
| 2010 | 17.0 | 18.0 | 16.0 | |
| 2011 | 16.3 | 15.7 | 17.0 | |
| Initial course of treatment | ||||
| Surgery | 6.3 | 8.1 | 4.4 | <0.001 |
| Radiation | 8.3 | 8.8 | 7.8 | 0.256 |
| Census tract median household income | ||||
| 1st quartile | 18.0 | 20.7 | 15.1 | |
| 2nd quartile | 20.8 | 22.0 | 19.6 | <0.001 |
| 3rd quartile | 25.4 | 25.3 | 25.6 | |
| 4th quartile | 35.7 | 32.0 | 39.7 | |
| Census tract % of residents living in poverty | ||||
| 1st quartile | 31.6 | 28.1 | 35.5 | |
| 2nd quartile | 27.7 | 27.3 | 28.0 | <0.001 |
| 3rd quartile | 22.4 | 24.1 | 20.7 | |
| 4th quartile | 18.3 | 20.6 | 15.8 | |
| Census tract % of residents 25 yrs or older with <12 years of education | ||||
| 1st quartile | 28.2 | 24.7 | 31.9 | |
| 2nd quartile | 26.8 | 26.2 | 27.4 | <0.001 |
| 3rd quartile | 23.2 | 25.2 | 21.0 | |
| 4th quartile | 21.9 | 23.9 | 19.7 |
Chemotherapy is 1 (0 otherwise) if patient initiated chemotherapy within 60 days of diagnosis; Numbers presented are percentages; p-values are based on Chi-square tests
Our propensity score matched sample achieved balance (i.e., ≤10% standardized difference)(29) in almost all patient characteristics that were found to be imbalanced in the original sample (eTable 2).
Using the matched sample, patients who received chemotherapy had a median survival of 122 days vs. 97 days among patients who did not receive chemotherapy (data not shown). Patients who initiated chemotherapy were much more likely to have received care in a hospital or ED in the last 30 days of life compared to patients without chemotherapy (Table 2): 44.7% vs. 29.5% had at least one hospital admission (p<0.001), 41.1% vs. 27.4% had at least one ED visit (p<0.001), and 14.1% vs. 9.2% died in a hospital (p<0.001). Although the probability of receiving at least one day of hospice care did not differ by chemotherapy use, patients receiving chemotherapy received a median of 8 days of hospice care vs. 15 days among patients not receiving chemotherapy (p<0.001). Median Medicare payment in the last 30 days of life were more than $3,000 greater ($12,338.7 vs. $8,907.6), and median patient OOP costs more than $750 greater ($1,004.8 vs. $228.5), for patients with vs. without chemotherapy (p<0.001 for both comparisons based on tests of the medians).
Table 2.
End-of-life heath care use and costs by whether patient initiated chemotherapy, using propensity score-matched sample
| Unadjusted Analysis
|
Adjusted Analysisb
|
||||||
|---|---|---|---|---|---|---|---|
| Chemo = 0 | Chemo = 1 | p-value | Chemo=0 | Chemo=1 | Difference (chemo vs. no-chemo) | p-value | |
| Any Inpatient Admission (%) | 29.5 | 44.7 | <0.001 | 29.2 (24.4, 34.1) |
45.0 (42.7, 47.3) |
15.7 (10.5, 21.0) |
<0.001 |
| Any ED visit (%) | 27.4 | 41.1 | <0.001 | 27.2 (22.5, 31.9) |
41.3 (39.0, 43.7) |
14.1 (9.0, 19.3) |
<0.001 |
| Any hospice use (%) | 75.6 | 75.6 | 0.969 | 75.7 (71.3, 80.1) |
75.5 (73.4, 77.5) |
−0.3 (−5.0, 4.5) |
0.916 |
| Died in a hospital (%) | 9.2 | 14.1 | <0.001 | 9.1 (6.3, 11.9) |
14.2 (12.6, 15.9) |
5.1 (2.0, 8.3) |
0.001 |
| # of days receiving hospicea | 15 | 8 | <0.001 | 15.7 (14.4, 17.0) |
11.5 (11.0, 12.0) |
−4.2 (−5.5, −2.8) |
<0.001 |
| Medicare payment ($) a | 8907.6 | 12338.7 | <0.001 | 15069.8 (13126.0, 17013.6) |
15951.2 (15175.6, 16726.8) |
881.4 (−953.2, 2716.0) |
0.346 |
| Patient OOP ($) a | 228.5 | 1004.8 | <0.001 | 841.0 (677.7, 1004.3) |
1311.5 (1229.5, 1393.4) |
470.5 (315.4, 625.6) |
<0.001 |
| Medicare + OOP payment ($)a | 9023.2 | 13649.1 | <0.001 | 15921.6 (13851.8, 17991.3) |
17244.5 (16420.7, 18068.2) |
1322.9 (−625.2, 3271.0) |
0.183 |
Both unadjusted and adjusted analyses were conducted on propensity score matched sample.
For unadjusted analysis, chi-square tests were performed for dichotomous outcomes; median tests were performed for continuous outcomes.
Numbers presented are medians for unadjusted results and means (95% confidence intervals) for adjusted results
Numbers presented are predicted outcomes associated with receiving chemotherapy within last 30 days of a Stage IV pancreatic cancer diagnosis
Each model adjusts for demographics, socioeconomic indicators at census tract level, number of co-morbidities at baseline, year of diagnosis, and other treatment (surgery or radiation) as part of first-course treatment
Adjusted analysis using the matched sample generated similar results as seen in unadjusted analysis (Table 2). For example, chemotherapy use was associated with a 16-percentage-point increase (95% confidence interval, 10.5, 21.0; p<0.001) in the probability of any hospital admission. Estimated increase in mean Medicare payments associated with chemotherapy use did not achieve statistical significance. However, chemotherapy was associated with a more than 50% increase in mean OOP costs (from $841.0 to $1311.5, p<0.001). Adjusted analysis using the entire sample but with inverse probability weighting generated very similar results (eTable 3).
Chemotherapy vs. no chemotherapy use in the last 30 days of life
Among patients who initiated chemotherapy within 60 days of diagnosis, patients who received chemotherapy in the last 30 days of life (N=749) (vs. who stopped chemotherapy earlier, N=1,091) were more likely to be male and less likely to have had radiation as part of first course treatment, but did not differ statistically along other dimensions (eTable 4).
Unadjusted analysis by receipt of chemotherapy in the last 30 days of life showed stark differences in health care utilization and costs at end-of-life (Table 3). Patients receiving chemotherapy in the last 30 days of life were almost twice as likely to have had a hospital admission or ED visit, less likely to have used hospice (61.4% vs. 84.7%, p<0.001), had much lower median days in hospice (3 vs. 16 days, p<0.001), and much more likely to have died in a hospital (23.9% vs. 7.5%, p<0.001). Medicare payment, patient OOP costs, and combined Medicare and patient payments were all significantly higher (p<0.001) among patients who received chemotherapy in the last 30 days of life. Adjusted analysis generated similar results (Table 3). Of note, chemotherapy use in the last 30 days of life was associated with a 28% increase in mean Medicare payment (from $14,032.8 to $17,959.6, p<0.001) and more than doubling of mean patient OOP costs (from $815.40 to $1,930.70, p<0.001).
Table 3.
End-of-life heath care use and costs among chemotherapy users by whether patient received chemotherapy in the last 30 days of life
| Unadjusted Analysis
|
Adjusted Analysisb
|
||||||
|---|---|---|---|---|---|---|---|
| Chemo in last 30 days = 0 | Chemo in last 30 days = 1 | p-value | Chemo in last 30 days = 0 | Chemo in last 30 days = 1 | Difference (chemo vs. no-chemo) | p-value | |
| Any Inpatient Admission (%) | 32.2 | 62.8 | <0.001 | 32.1 | 62.9 | 30.8 | <0.001 |
| Any ED visit (%) | 30.9 | 55.9 | <0.001 | 30.9 | 55.8 | 24.9 | <0.001 |
| Any hospice use (%) | 84.7 | 61.4 | <0.001 | 84.8 | 62.3 | −22.6 | <0.001 |
| Died in a hospital (%) | 7.5 | 23.9 | <0.001 | 7.3 | 23.6 | 16.3 | <0.001 |
| # of days receiving hospicea | 16 | 3 | <0.001 | 16 (15.5, 16.4) |
5.1 (4.8, 5.4) |
−10.9 (−11.4, −10.3) |
<0.001 |
| Medicare payment($) a | 9109.9 | 16376.1 | <0.001 | 14032.8 (13273.7, 14791.9) |
17959.6 (16784.6, 19134.6) |
3926.8 (2531.5, 5322.2) |
<0.001 |
| Patient OOP($) a | 325.2 | 1819.4 | <0.001 | 815.4 (759.2, 871.6) |
1930.7 (1769.0, 2092.5) |
1115.3 (944.2, 1286.5) |
<0.001 |
| Medicare + OOP payment($)a | 9499.0 | 18508.3 | <0.001 | 14848.0 (14060.1, 15635.9) |
19884.1 (18607.6, 21160.6) |
5036.1 (3539.3, 6533.0) |
<0.001 |
No propensity score adjustment was conducted for either unadjusted or adjusted analysis.
For unadjusted analysis, chi-square tests were performed for dichotomous outcomes; median tests were performed for continuous outcomes.
Numbers presented are medians for unadjusted results and means (95% confidence intervals) for adjusted results
Numbers presented here are predicted outcomes associated with receiving chemo in the last 30 days of life
Each model adjusts for demographics, socioeconomic indicators at census tract level, number of co-morbidities at baseline, year of diagnosis, and other treatment (surgery or radiation) as part of first-course treatment
Discussion
Chemotherapy use among older patients diagnosed with Stage IV pancreatic cancer was associated increased probabilities of hospitalizations, ED visits, and hospital death, fewer hospice days, and substantial (great than 50%) increase in patient out-of-pocket costs for care in the last 30 days of life. More stark differences in these outcomes were found between patients who received chemotherapy in the last 30 days of life and those who initiated chemotherapy but did not continue to receive chemotherapy in the last 30 days, including more than a two-fold increase in mean patient OOP costs for care in the last month of life.
Our findings highlight the potential harm of chemotherapy for quality-of-life close to death and thus add to the growing concern regarding the value of chemotherapy for patients with metastatic cancer and limited life-expectancy.(3,4,30,31) Although our data do not allow us to measure QOD directly, increased hospitalizations, ED visits, and deaths in a hospital all predict worse QOD.(32) These findings highlight that harms of chemotherapy use among dying cancer patients often exceed its benefits.
Our findings provide much needed data to support clinician-patient discussion when making decisions about chemotherapy for metastatic pancreatic cancer. Discussion of prognosis with one’s oncologist is associated with a better patient understanding of the terminal nature of their illness.(33) However, many patients with incurable metastatic cancer initiate chemotherapy with unrealistic expectations of its potential benefits.(34) A substantial minority of patients with advanced-stage cancer and an oncologist-estimated life expectancy of 6 months or less never had a discussion about life-expectancy or prognosis with their oncologists.(33) The data we generate on what chemotherapy could mean for patients’ quality of life in the months ahead will better inform the individual decision on whether to initiate chemotherapy.
Our findings also highlight the importance of continuously reevaluating the value of chemotherapy among patients who initiated it. In our study sample, more than 40% of Stage IV pancreatic cancer patients who initiated chemotherapy received chemotherapy in the last 30 days of life. Ninety-five percent of patients who initiated chemotherapy survived until at least 30 days after initiation and close to 80% survived until at least 60 days after initiation (data not shown). This suggests that many patients who continued to receive chemotherapy in the last 30 days might have had a chance to stop chemotherapy earlier but did not. While termination of chemotherapy is a difficult decision in the context of substantial uncertainty regarding the benefits and harms of continued chemotherapy, our findings add to existing knowledge of potential detriments to QOD of chemotherapy use very close to EOL, which, in turn, will assist continued monitoring and discussion by oncologists and their patients.
One unique finding of our study concerns the substantial increase in patient OOP costs associated with chemotherapy use and continued chemotherapy use close to end-of-life. Financial toxicity, including financial burdens of cancer treatment in terms of OOP costs, depletion of assets, medical debt, bankruptcy, and related stress have received increasing attention from the National Cancer Institute.(35) Previous studies report higher annual OOP costs among adults recently diagnosed with cancer than among adults without a history of cancer.(36) We found last-month-of-life OOP costs for older patients diagnosed with advanced pancreatic cancer to be comparable to published estimates of the annual OOP costs for non-elderly cancer survivors.(36) More importantly, initiation of chemotherapy in this patient population further added to the financial burden by increasing last-month-of-life OOP costs by more than 50%; the incremental burdens associated with continued use of chemotherapy very close was even greater. Our findings thus bring to light the substantial financial burdens (and, likely, distress) to patients and families associated with chemotherapy for older patients with advanced cancer despite comprehensive coverage of cancer treatment by Medicare. This information, coupled with the QOD impairment associated with chemotherapy use, ought to be part of the communication to patients and their families as they make the difficult decisions regarding care for advanced cancer with limited life-expectancy.(35,37)
Our study has a few important limitations. In absence of data on patient baseline performance, prognostic information, and preferences, which prominently influence the clinical decision on chemotherapy,(38) we restricted the analytic sample to mitigate biases, following a previous study.(1) Specifically, we excluded those who died within 60 days of diagnosis (accounting for 44% of the original sample) and only considered chemotherapy that was initiated within 60 days of diagnosis (accounting for 75% of all patients who survived until at least 60 days and ever used chemotherapy). Our findings thus apply to the sub-population with a relatively better prognosis at baseline for whom chemotherapy presents as a realistic option. It is plausible that, even with such sample restrictions and propensity score adjustment, there were unobserved differences between individuals who did and did not use chemotherapy (and between those who did or did not use it in the last 30 days of life) that would bias our estimates. However, such residual confounding (e.g., by baseline performance) would bias our results in the direction of suggesting smaller differences in health care use and costs close to end of life, leading to conservative estimates.
Another limitation concerns our retrospective design. To examine health care use and costs at the end of life, we had to restrict to a sample of patients who died within 12 month/360 days after diagnosis. The biases introduced by such a retrospective design(39) were mitigated by the fact that 90% of our sample diagnosed with metastatic pancreatic cancer died within 12 months of diagnosis. Additionally, some of the health care events we examined (e.g., ED visits) might have preceded chemotherapy use – another reason that our results should be interpreted as associations, although a very small proportion (5%) of patients who received chemotherapy in our sample died within 30 days after initiation of chemotherapy.
Conclusions
Chemotherapy use among older patients diagnosed with metastatic pancreatic cancer was associated with a higher likelihood of hospitalization and ED use, fewer days of hospice care, and substantially increased patient out-of-pocket costs, in the last 30 days of life. When clinicians discuss initiating and continuing chemotherapy with patients with advanced cancer, associations among chemotherapy, QOD, and heightened financial burdens should be taken into account.
Supplementary Material
Acknowledgments
Funding: This work was supported by Weill Cornell Medical College; the National Cancer Institute [grant number 1R35CA197730]; and the Department of Veterans Affairs [grant number CDP 12–255].
Role of the Funder/Sponsor: Sponsors of this research played no role in the design and conduct of the study; collection, management, analysis, and interpretation of the data; preparation, review, or approval of the manuscript; and decision to submit the manuscript for publication.
Footnotes
Publisher's Disclaimer: This is a PDF file of an unedited manuscript that has been accepted for publication. As a service to our customers we are providing this early version of the manuscript. The manuscript will undergo copyediting, typesetting, and review of the resulting proof before it is published in its final citable form. Please note that during the production process errors may be discovered which could affect the content, and all legal disclaimers that apply to the journal pertain.
Conflicts of Interest Disclosure: The authors report no conflicts of interest.
Disclaimer: The views expressed in this article are those of the authors and do not necessarily reflect the policy or position of the Department of Veterans Affairs or the United States government.
Results from an early phase of the study (reporting on a subset of the outcomes examined in this manuscript) was presented as a poster at the 2016 ASCO annual meeting.
References
- 1.Earle CC, Tsai JS, Gelber RD, et al. Effectiveness of chemotherapy for advanced lung cancer in the elderly: instrumental variable and propensity analysis. J Clin Oncol. 2001;19(4):1064–1070. doi: 10.1200/JCO.2001.19.4.1064. [DOI] [PubMed] [Google Scholar]
- 2.NSCLC Meta-Analyses Collaborative Group. Chemotherapy in addition to supportive care improves survival in advanced non-small-cell lung cancer: a systematic review and meta-analysis of individual patient data from 16 randomized controlled trials. J Clin Oncol. 2008;26(28):4617–4625. doi: 10.1200/JCO.2008.17.7162. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 3.Wright AA, Zhang B, Keating NL, Weeks JC, Prigerson HG. Associations between palliative chemotherapy and adult cancer patients’ end of life care and place of death: prospective cohort study. BMJ. 2014;348 doi: 10.1136/bmj.g1219. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 4.Prigerson HG, Bao Y, Shah MA, et al. Chemotherapy Use, Performance Status, and Quality of Life at the End of Life. JAMA Oncol. 2015 doi: 10.1001/jamaoncol.2015.2378. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 5.Ersek M, Miller SC, Wagner TH, et al. Association between aggressive care and bereaved families’ evaluation of end-of-life care for veterans with non-small cell lung cancer who died in Veterans Affairs facilities. Cancer. 2017 doi: 10.1002/cncr.30700. [DOI] [PubMed] [Google Scholar]
- 6.Lubitz JD, Riley GF. Trends in Medicare Payments in the Last Year of Life. N Engl J Med. 1993;328(15):1092–1096. doi: 10.1056/NEJM199304153281506. [DOI] [PubMed] [Google Scholar]
- 7.Riley GF, Lubitz JD. Long-Term Trends in Medicare Payments in the Last Year of Life. Health Serv Res. 2010;45(2):565–576. doi: 10.1111/j.1475-6773.2010.01082.x. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 8.Riley GF. Long-Term Trends In The Concentration Of Medicare Spending. Health Aff. 2007;26(3):808–816. doi: 10.1377/hlthaff.26.3.808. [DOI] [PubMed] [Google Scholar]
- 9.Ramsey S, Blough D, Kirchhoff A, et al. Washington State cancer patients found to be at greater risk for bankruptcy than people without a cancer diagnosis. Health Aff (Millwood) 2013;32(6):1143–1152. doi: 10.1377/hlthaff.2012.1263. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 10.Shankaran V, Jolly S, Blough D, Ramsey SD. Risk factors for financial hardship in patients receiving adjuvant chemotherapy for colon cancer: a population-based exploratory analysis. J Clin Oncol. 2012;30(14):1608–1614. doi: 10.1200/JCO.2011.37.9511. [DOI] [PubMed] [Google Scholar]
- 11.Shankaran V, Ramsey S. Addressing the financial burden of cancer treatment: From copay to can’t pay. JAMA Oncol. 2015;1(3):273–274. doi: 10.1001/jamaoncol.2015.0423. [DOI] [PubMed] [Google Scholar]
- 12.Zafar SY, McNeil RB, Thomas CM, et al. Population-based assessment of cancer survivors’ financial burden and quality of life: a prospective cohort study. J Oncol Pract. 2015;11(2):145–150. doi: 10.1200/JOP.2014.001542. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 13.Muniraj T, Jamidar PA, Aslanian HR. Pancreatic cancer: a comprehensive review and update. Dis Mon. 2013;59(11):368–402. doi: 10.1016/j.disamonth.2013.08.001. [DOI] [PubMed] [Google Scholar]
- 14.Conroy T, Desseigne F, Ychou M, et al. FOLFIRINOX versus Gemcitabine for Metastatic Pancreatic Cancer. N Engl J Med. 2011;364(19):1817–1825. doi: 10.1056/NEJMoa1011923. [DOI] [PubMed] [Google Scholar]
- 15.Von Hoff DD, Ervin T, Arena FP, et al. Increased Survival in Pancreatic Cancer with nab-Paclitaxel plus Gemcitabine. N Engl J Med. 2013;369(18):1691–1703. doi: 10.1056/NEJMoa1304369. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 16.Tsavaris N, Kosmas C, Skopelitis H, et al. Second-line treatment with oxaliplatin, leucovorin and 5-fluorouracil in gemcitabine-pretreated advanced pancreatic cancer: A phase II study. Invest New Drugs. 2005;23(4):369–375. doi: 10.1007/s10637-005-1446-y. [DOI] [PubMed] [Google Scholar]
- 17.Gilabert M, Chanez B, Rho YS, et al. Evaluation of gemcitabine efficacy after the FOLFIRINOX regimen in patients with advanced pancreatic adenocarcinoma. Medicine (Baltimore) 2017;96(16) doi: 10.1097/MD.0000000000006544. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 18.National Cancer Institute: Division of Cancer Control & Population Sciences. Procedure Codes for SEER-Medicare Analyses. https://healthcaredelivery.cancer.gov/seermedicare/considerations/procedure_codes.html. Accessed June 6, 2017.
- 19.U.S. Department of Labor: Bureau of Labor Statistics. CPI Inflataion Calculator. https://www.bls.gov/data/inflation_calculator.htm. Accessed June 6, 2017.
- 20.Warren JL, Butler EN, Stevens J, et al. Receipt of Chemotherapy Among Medicare Patients With Cancer by Type of Supplemental Insurance. J Clin Oncol. 2015;33(4):312–318. doi: 10.1200/JCO.2014.55.3107. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 21.Stedman MR, Doria-Rose P, Warren JL, Klabunde CN, Mariotto A. Comorbidity Technical Report: The Impact of Different SEER-Medicare claims-based Comorbidity Indexes on Predicting Non-cancer Mortality for Cancer Patients. https://healthcaredelivery.cancer.gov/seermedicare/considerations/comorbidity-report.html. Accessed June 6, 2017.
- 22.Garrido MM, Kelley AS, Paris J, et al. Methods for constructing and assessing propensity scores. Health Serv Res. 2014;49(5):1701–1720. doi: 10.1111/1475-6773.12182. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 23.Austin PC. Optimal caliper widths for propensity-score matching when estimating differences in means and differences in proportions in observational studies. Pharm Stat. 2011;10(2):150–161. doi: 10.1002/pst.433. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 24.Ho DE, Imai K, King G, Stuart EA. Matching as Nonparametric Preprocessing for Reducing Model Dependence in Parametric Causal Inference. Polit Anal. 2007;15(3):199–236. [Google Scholar]
- 25.Long JS. Regression models for categorical and limited dependent variables. Thousand Oaks, CA: SAGE Pubications, Inc.; 1997. [Google Scholar]
- 26.Manning WG, Basu A, Mullahy J. Generalized modeling approaches to risk adjustment of skewed outcomes data. J Health Econ. 2005;24(3):465–488. doi: 10.1016/j.jhealeco.2004.09.011. [DOI] [PubMed] [Google Scholar]
- 27.Stuart EA. Matching methods for causal inference: A review and a look forward. Stat Sci. 2010;25(1):1–21. doi: 10.1214/09-STS313. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 28.Abadie A, Imbens GW. Matching on the Estimated Propensity Score. Econometrica. 2016;84(2):781–807. [Google Scholar]
- 29.Austin PC. Balance diagnostics for comparing the distribution of baseline covariates between treatment groups in propensity-score matched samples. Stat Med. 2009;28(25):3083–3107. doi: 10.1002/sim.3697. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 30.Nilsen ML, Johnson JT. Potential for low-value palliative care of patients with recurrent head and neck cancer. Lancet Oncol. 2017;18(5):e284–e289. doi: 10.1016/S1470-2045(17)30260-7. [DOI] [PubMed] [Google Scholar]
- 31.Roeland E, Loprinzi C, Moynihan TJ, Smith TJ, Temel J. In chemotherapy for lung cancer, sometimes less is more. J Natl Compr Canc Netw. 2013;11(3):232–235. doi: 10.6004/jnccn.2013.0034. [DOI] [PubMed] [Google Scholar]
- 32.Zhang B, Nilsson ME, Prigerson HG. Factors important to patients’ quality of life at the end of life. Arch Intern Med. 2012;172(15):1133–1142. doi: 10.1001/archinternmed.2012.2364. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 33.Epstein AS, Prigerson HG, O’Reilly EM, Maciejewski PK. Discussions of Life Expectancy and Changes in Illness Understanding in Patients With Advanced Cancer. J Clin Oncol. 2016;34(20):2398–2403. doi: 10.1200/JCO.2015.63.6696. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 34.Weeks JC, Catalano PJ, Cronin A, et al. Patients’ Expectations about Effects of Chemotherapy for Advanced Cancer. N Engl J Med. 2012;367(17):1616–1625. doi: 10.1056/NEJMoa1204410. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 35.PDQ® Adult Treatment Editorial Board. PDQ Financial Toxicity and Cancer Treatment. 2017 https://www.cancer.gov/about-cancer/managing-care/financial-toxicity-hp-pdq. Accessed June, 6 2017.
- 36.Guy GP, Jr, Ekwueme DU, Yabroff KR, et al. Economic Burden of Cancer Survivorship Among Adults in the United States. J Clin Oncol. 2013;31(30):3749–3757. doi: 10.1200/JCO.2013.49.1241. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 37.Zafar SY. Financial Toxicity of Cancer Care: It’s Time to Intervene. J Natl Cancer Inst. 2016;108(5) doi: 10.1093/jnci/djv370. [DOI] [PubMed] [Google Scholar]
- 38.ECOG-ACRIN Cancer Research Group. ECOG Performance Status. http://ecog-acrin.org/resources/ecog-performance-status. Accessed June 3, 2015.
- 39.Bach PB, Schrag D, Begg CB. Resurrecting treatment histories of dead patients: A study design that should be laid to rest. JAMA. 2004;292(22):2765–2770. doi: 10.1001/jama.292.22.2765. [DOI] [PubMed] [Google Scholar]
Associated Data
This section collects any data citations, data availability statements, or supplementary materials included in this article.